search

microsoft / nnfusion

A flexible and efficient deep neural network (DNN) compiler that generates high-performance executable from a DNN model description.
MIT License
937 stars 157 forks source link

Invoke all ONNX tests #502

Closed wenxcs closed 1 year ago

wenxcs commented 1 year ago 
$ ./test/python/run_onnx_test.sh
**will generate:**
# wenxh @ srgam-02 in ~/nnfusion_onnxtest/nnfusion/test_snapshot on git:xbox_test x [12:58:39] C:130
$ ls
default.csv  float16.csv  float64.csv  input_as_constant.csv

**Could be compared to ground truth in:**
# wenxh @ srgam-02 in ~/nnfusion_onnxtest/nnfusion/test/python/ground_truth on git:xbox_test x [12:59:15]
$ ls
default.csv  float16.csv  float64.csv  input_as_constant.csv

# wenxh @ srgam-02 in ~/nnfusion_onnxtest/nnfusion on git:xbox_test x [14:56:37] 
$ ./test/python/run_onnx_test.sh
Bad cases:
FP32/Input as parameter:
@, add , test_add , BUILD ERROR , FAILED
vs ground truth:
@, add , test_add ,/home/wenxh/nnfusion_onnxtest/nnfusion/test/python/../../build/onnx/onnx/backend/test/data/node/test_add/test_data_set_0 , PASS

FP32/Input as constant:
FP16/Input as constant:
FP64/Input as constant:

# wenxh @ srgam-02 in ~/nnfusion_onnxtest/nnfusion on git:xbox_test x [12:49:13] 
$ ./test/python/run_user_list_compare.sh 
Error: test_split_equal_parts_1d_opset13 with config[ default,float16,float64,input_as_constant ] is not FAILED
Error: test_convtranspose_kernel_shape with config[ default,float16,float64,input_as_constant ] is not FAILED
Error: test_cast_FLOAT_to_BFLOAT16 with config[ default,float16,float64,input_as_constant ] is not FAILED
Error: test_cast_FLOAT_to_STRING with config[ default,float16,float64,input_as_constant ] is not FAILED